Daisytuner

Daisytuner operates as an AI-powered co-pilot designed to generate high-performance code by integrating into a client's existing build pipeline. The service addresses the limitations of standard compilers by applying advanced, data-centric optimization techniques.

The company's solution is targeted at developers and organizations in computationally intensive fields, including numerical simulations, artificial intelligence, and edge computing. For AI applications, the technology aims to reduce the monetary costs associated with training and inference. In edge computing, it helps meet real-time processing requirements on devices with limited resources.

The business model is centered around providing this optimization tool, which leverages the DaCe (Data-Centric Parallel Programming) framework. It functions by analyzing code, benchmarking the target processor architecture using tools like Likwid, and applying transformations based on similarity from cloud databases to optimally utilize heterogeneous hardware. This process results in a C library that can be called from Python, enhancing performance for specific computational tasks.

Keywords: code optimization, high-performance computing, AI co-pilot, numerical simulation, edge computing, compiler, data-centric optimization, DaCe framework, performance tuning, heterogeneous architectures